Hi all, I have 12 open PRs that could use a review. If you're looking for a smaller starting point, #18694 (JDBC properties), #18691 (Hive view protection), and #18695 (metrics test assertions) are focused fixes.
I've grouped the full list by area below. CI is green on the current heads, and GitHub currently reports no merge conflicts. *Core / API* - #18683 <https://github.com/apache/iceberg/pull/18683>: Make lazy initialization in BaseSnapshot and Schema thread-safe. - #18684 <https://github.com/apache/iceberg/pull/18684>: Support empty struct defaults in Java and the REST spec, preserving defaults during column pruning. - #18694 <https://github.com/apache/iceberg/pull/18694>: Keep Iceberg-specific JDBC catalog properties out of driver connection properties. - #18591 <https://github.com/apache/iceberg/pull/18591>: Send snapshot events and commit reports after transaction commit. The event-clearing, ordering-test and documentation feedback has been addressed. *Hive* - #18691 <https://github.com/apache/iceberg/pull/18691>: Prevent dropTable from dropping an Iceberg view. - #18692 <https://github.com/apache/iceberg/pull/18692>: Check view commit status after an HMS conflict error to avoid deleting metadata for a successful commit. *AWS / Snowflake* - #18693 <https://github.com/apache/iceberg/pull/18693>: Apply S3 access grants and user-agent configuration in AssumeRoleAwsClientFactory. - #18589 <https://github.com/apache/iceberg/pull/18589>: Keep analytics accelerator factories open while their FileIOs are still in use. Feedback on the cleanup approach would be helpful. - #18690 <https://github.com/apache/iceberg/pull/18690>: Use FileIOTracker to avoid retaining every FileIO until SnowflakeCatalog closes. *Spark / Kafka Connect / Tests* - #18494 <https://github.com/apache/iceberg/pull/18494>: Avoid scanning every manifest for an unbounded sync streaming read. The merged-manifest offset fix and regression test are included. - #18687 <https://github.com/apache/iceberg/pull/18687>: Stop the Kafka Connect coordinator when its leader partition is revoked. - #18695 <https://github.com/apache/iceberg/pull/18695>: Fix writer metrics assertions to check both bounds for every column beyond the inferred-column limit. Thanks to everyone who has already taken a look. I'd appreciate another look at the updates on #18494 and #18591, and guidance on the cleanup approach for #18589. If you have context in any of these areas, a review or a pointer to the right reviewer would help. Test details are in each PR. Thanks, Daksh GitHub: binarybrain-009
